Letters

AI under review

6 November 2024

Dear Editors,

I am very concerned that Paul Cleverley and colleagues have chosen to publish the article, Advancing transparent and ethical AI, at this time and am surprised to see it published. The article repeats arguments already presented in the autumn edition of Geoscientist (Cleverley, 2024) and addressed by Stephenson et al. (2024a, 2024b). More seriously the article transgresses the International Union of Geological Sciences’ (IUGS) ongoing review processes.

The authors were part of the IUGS-led discussion of Large Language Models (LLM) held at the Geological Society in July (a summary of which is publicly available under Chatham House Rule; IUGS, 2024), which identified clear next steps. The authors are aware, as discussed at the IUGS Council meeting held at the International Geological Congress in Busan in August 2024, that there is an ongoing ad-hoc review of the DDE activity by IUGS – indeed the authors refer to this review in their article.

The review is a process in IUGS in which a root-and-branch evaluation of the scientific value of an IUGS activity is assessed, along with its governance and the effectiveness of its structure. The article undermines the review process and the development of trust that was building with regards to the development of a LLM in the geosciences (which associates the entities of DDE and GeoGPT). At best this could be considered a breach of best practice, but also poses questions on contravention of the ethics related to the DDE review and LLM discussion process.

The article should be retracted pending the outcome of the review and ongoing discussion with the owners of the draft GeoGPT LLM. We expect the review process to be complete by the end of 2024. The outcomes of the review, the updated terms of reference for the DDE project, and the relationship between DDE and the owners of GeoGPT will be presented to the IUGS Executive at their February 2025 meeting.

Prof John Ludden (On behalf of the IUGS Executive)

Past-President of the IUGS, Chair of the ad-hoc review committee of IUGS for the DDE project, and organiser of a recent IUGS discussion on LLM.

(Letter received 12 Oct 2024)

Response:

The Geoscientist Editorial Team (Ruth Allington, Editor-in-Chief (incoming), Prof Andy Fleet, Editor-in-Chief (outgoing); David Shilston, Deputy Editor-in-Chief; Dr Amy Whitchurch, Executive Editor; Dr Marissa Lo, Associate Editor), write in response:

We thank John Ludden for his letter and welcome his views and feedback. Geoscientist is editorially independent from the Society and Council, so decisions to publish and retract content are made on editorial grounds. In this case, we could not find any grounds to retract the article, Advancing Transparent and Ethical AI, from Paul Cleverley and colleagues.

The article is part of ongoing discussions between the authors and the DDE team about AI in the geosciences, and specifically about DDE’s geoscience ‘chatbot’ GeoGPT (Cleverley, 2024; Stephenson et al., 2024a, 2024b). Many of the issues raised during these discussions, which began in early 2024, remain unresolved, are relevant to the broader geoscience community and of public interest.

As far as we are aware, there is no formal embargo around the IUGS-led review of DDE, which was initiated earlier this year and is noted in the IUGS summary of the meeting held at Burlington House in July 2024 (and released under Chatham House Rule; IUGS, 2024). If restrictions on activities and reporting are in place, we assume this would also extend to the DDE team working on GeoGPT, yet GeoGPT continues to be extensively promoted and demonstrated, in these pages and beyond, not least during the IUGS International Geoscience Congress in August and the Geological Society of America Connects meeting in September.

AI is developing rapidly and we believe it is in the interest of the whole geoscience community for issues around AI to be openly and constructively debated in the timeliest manner possible. As an editorially independent Fellowship magazine, we aim to provide an impartial forum for debate and welcome contributions on this topic from all sides and stakeholders.

Further reading